Assignment

Public school teachers in the U.S. are almost all paid according to rigid pay scales known as "steps and lanes." Teachers' salaries go up by a certain percentage with each additional year on the job (a "step"), and they move to a different "lane" with some percentage jump in pay if they have or earn a graduate degree. On the surface, because of this rigidity it seems unlikely that there would be a gender gap in pay. Your job is to ftnd evidence about whether that is true by using regressions.

The complete NSCG 2015 data file (nscg15.rdata) is available on Google Drive. The codebook and questionnaire are also available and will be needed.

Expectations about content

1. I encourage you to work together in thinking through the data preparation and regression specifications. However, as always, you need to do your own writing.

2. Write this up as though it is part of a (short) research paper.

- You don't need to write an introduction or look up relevant literature or any of that stuff. Expected total text ≈ 2-3 pages. Use R Markdown, as usual.

- Do explain the research question, i.e., what you're trying to do.

- Precisely explain the variables you use. You've already worked on identifying public school teachers in a previous assignment, but you need to explain in words whom you considered a public school teacher.

The NSCG data allow you to find time in the current job, so don't assume people have had the same job since they finished school. Time in a job is usually referred to as "tenure," but that doesn't mean tenure in the sense of job security.

- Explain your regression specifications. The usual way to do this is to first explain a baseline specification then explain why you estimated variations on that.

- Evaluate the results, both in statistical and practical terms.

3. It's important to incorporate the information above about steps and lanes into your regression model, but there isn't a single correct way to do it. That's why people usually report more than one regression in empirical research.

4. Report your results in a nice ( tabLego) table or tables.

Some other things to consider for your regressions and/or interpretation

1. Definitely think about interactions (hint, hint). If you use them, explain why.

2. Some additional factors you could think about: elementary vs. secondary, teaching field (e.g., social studies), region.

3. Other nonlinearities (like quadratic terms). If you use them, explain why.

4. Appropriate hypothesis tests, especially regarding the research question.

Expectations for tables and writing

Although what you produce will be short, it should meet professional for communication. Your grade will include a separate communications score.

1. Proofread your assignment before turning it in, preferably more than once.

2. Tables must be numbered (even if there is only one) and have titles and appropriate notes. When you discuss your results, refer to the table by number. Charts are not expected, but if you get creative and use one, the same rule applies.

3. Use ordinary English in text and tables. No mysterious R or NSCG variable names allowed! Except in your code blocks, everything in the assignment should be in ordinary English.1

4. Think about the appropriate number of decimal places for your table and numbers you refer to in the text. Too few and many results might round to zero. Too many makes it hard to read.

5. Be sure to round, not truncate. For example, 20.46 rounds to 20.5, not 20.4.

6. Remove all unneeded R code and output. For example, don't print all of your regressions with summary(lm(...)) and then show the tabLego table. The rule is that if you don't mention it, don't show it.

7. Did I mention proofreading?